levels for delivering stems

[Atardecer]

Hey all,

Am curious as to how you set your levels when you're doing music for film and you have to deliver as stems. So do you mix to the dialogue, which could result in quite low levels in certain cues, or do you basically just get as much juice as you can regardless of how loud you want a particular cue to be overall (keeping your mix balance obviously), and just let the dubbing guys deal with it. I had been mixing mainly to dialogue but am concerned about the SNR of some cues. Also the temp dialogue track i received is just that; a temp cut and is far too loud in certain parts so it may not be an accurate guide. Obviously i want to retain the dynamics and im avoiding any heavy compression or limiting (its mostly classical bases music anyway) but am in two minds as which way to go. Any thoughts?

James

[Ginger]

This used to be a bigger deal when dub stages were still analog because you didn't want the dubbing mixer to have to run the music faders down around -20 or so if you printed a hot-as-possible mix. I don't know how things are over where you are, but in LA almost all the dub stages are digital and usually the music mix is printed nice and hot (without going over of course) and then the dubbing mixer can run the faders where ever they need to. However, mixing to dialogue as far as making sure your mix is working with the dialog is critical or you will just end up getting all the music turned down during the dub. The composer is key in this too in that if they are good, they will have written out of the way of the dialog already. Meaning arrangement, instrumentation, dynamics will have been written with consideration for the dialog. That's the best way to help keep your score nice and loud in the dub.
When in doubt, call your dubbing mixer and ask what they want and/or expect. It's a good courtesy to get into the habit of too.

[PTUser NYC]

You know, the worst part about delivering stems is that if there is a compressor on the master fader, it reacts differently when you print the stems from how it would when the whole mix is playing.

I wonder if there is a possible feature here. I'm not wuite sure how it would be implemented, but the rough idea is that in some mode, the sidechain of the compressor on the master fader would be reacting to everything, while only the stem info would actually be being processed. So even if you were making a guitar stem, everytime the kick drum hit, the compressor would react anyway. In this way, if you added all the stems together, you realyl would get the same result as the overall mix.

I guess that its possible to do that now - you'd just put a post fader aux send from every channel, sending at unity, and then use that aux bus to sidechain the compressor. Then instead of muting channels you weren't including in the stripe, you'd have to assign them to an unused output so they wouldn't be included in the stem, but the aux send would still be active.

Wouldn't it be cool if there was some easier, automatic way to do this? Anyone making stems should really appreciate something like this, and yet I've never heard of anyone working like this.

[1m1]

The way I've always seen it done is compression on the stems, no compression on the whole mix.

[emilano]

Hey PTU, that is a great idea! Would be interesting to do a phase cancelation test on that and see if it actually was doing what we think it would...

I think you'd have to do it with a duplicated copy of the master compressor for each stem you wanted to make if you wanted to make them all at the same time right? Or you'd have to make your master mix and your stems by doing a separate pass for each stem.

Also I forget whether there is a slight delay when sending from a post fader send. And is there a delay in the sidechain input vs regular input? This could cause a slight difference in how the compressor reacts vs when it get its directly.

Starts getting complicated and making my brain hurt without trying it out, but I think its totally doable.

Oop. One problem I just realized. Most of the limiting plugs that I know of that are used as the final insert do not have side chain inputs (maxim, L1, L2007).

[emilano]

since we've gone off topic and because this is really a post production issue, i moved the last two posts to the pprod forum...

[MixerGuy]

The way I've always seen it done is compression on the stems, no compression on the whole mix.

wrong.

the stems MUST add up to equal the whole mix. Identical. End of story.

[PTUser NYC]

wrong.

the stems MUST add up to equal the whole mix. Identical. End of story.

Hey MixerGuy, I think what 1m1 is saying is that he does all his compression on submixes that becom the stems, rather than having a compressor across the mix bus precisely because then the stems WOULD equal the full mix.

Its the other way, where you have a compressor across the mix bus that produces different results as the compressor behaves differently processing the stems from when it runs the whole mix.

I'm not thinking that you don't know this already, I know you do. I'm just explaining the way you misread his post?

*I'M* the one doing this wrong, and I was wishing for a feature or technique that would allow me to have a compressor across the mix bus, while still being able to print stems that DO create an identical mix to the full mix.

The sidechain trick I outlined is the best I've come up with so far, but I'd love to see something a little easier to deal with.

[emilano]

The problem with compressing the stems only is that there's no guarantee that the sum of the stems won't go over whatever ceiling there needs to be on your final output (-10db for broadcast for example). Sure you could make all the stems peak on the low side to be safe, but that defeats the purpose here... You may want to be pushing things as hard as possible, or you may not want to have to work that hard. Not to mention the fact that a compressor on everything is much more cpu and tdm intensive.

[PTUser NYC]

i think the idea that 1m1 was putting forth, is that each stem be a submix, with a compressor on it, and that they are all audible as you create your mix. You can note the output level of that mix and stay under your spec, whatever that may be. But this is the only way that stems will add up to create the same final mix.

Obviously, if you have a compressor across the mix bus (and I usually do) then it hits the stems differently than the entire mix, and the sum of the stems will not add up to the same thing as the full mix.

I'd love to hear tips and techniques about how people deal with this. Its long been a question of mine.